minimisation_steps, an argument to feptasks.minimize() , currently matches the openmm LocalEnergyMinimizer.minimize() argument maxIterations in that when its equal to 0, it will run as many steps of minimization as needed, not zero steps.

minimisation_steps should work like this instead:

• None should be interpreted to run until tolerance is reached
• 0 should be interpreted to run 0 steps

See thread: #1065 (comment)

related: #1083